Everyone (including us) is asking: How does my brand show up in ChatGPT, Claude, and the other LLMs that are now used like Google Search? How do you measure AI search success?
So, we decided to test it for ourselves.
We learned the obvious: the entire concept of benchmarking your ranking in AI search is still new, and traditional search is still dominant. Optimizing AI search as a single system is, to put it lightly, impossible. Each platform has distinct preferences, citation logic, and trust hierarchies. Strategies must be tailored to individual platforms, aligning with their specific infrastructure, behaviors, and ranking priorities. Overhauling your entire content or communications strategy on GEO alone would be a big mistake, as 95% of Americans still rely on traditional search every day. But for how long?
The story of GEO
Some people see AI search and Generative Engine Optimization (GEO) as the evolution of SEO. Others see them as entirely separate. We see GEO as the spiritual descendant of traditional SEO, rooted in the same core principles of discoverability, relevance, and authority, but with unique dynamics we’re still getting familiar with.
Actively experimenting can give you a competitive edge as AI search continues to evolve. Early adopters who experiment with building visibility and credibility inside these systems will likely lead the next wave of digital branding, just as SEO adopters did during Google’s great rise in the early to mid-2000s.
We didn’t go into this experiment with high expectations, but we did come out of it with a better understanding of the existing tools to help measure GEO performance, and that improving your visibility in LLMs comes down to a standard comms philosophy: ensure your message is consistent, in as many places as possible, as frequently as possible.
Meet the tools that help us benchmark
Let’s start with what we’re actually trying to measure.
We wanted to understand why Inkhouse appeared in some searches (“What are some great PR firms for tech startups in the Bay Area?”) but not in others (“Who is Databricks’ PR agency of record?”). How could we ensure that Inkhouse’s name appears in every relevant search that we want to be included in, across all of the relevant LLMs? And what tools could help us figure this out?
LLMRefs was a solid starter tool. It helped us track and monitor how often Inkhouse appeared in AI search responses, helping us understand which prompts we were already showing up in and where gaps remained.
Cognizo is a full-stack GEO platform that, yes, is centered on AI search visibility, but provides deep analytics by testing prompts multiple times a day, benchmarking against competitors, and offering recommendations for actionable optimization.
We are also excited about the future of SEMRush’s AI traffic dashboard as our preferred SEO and digital marketing platform, but found the AI Traffic feature to still be in too early stages for this specific experiment.
But LLMRefs and Cognizo reiterated to us a central theme in comms and marketing: SEO and GEO (how we format content, where content resides, and how it’s optimized) are just layers on top of a strategic, compelling, and audience-centric story.
TL;DR: Write for humans; optimize for engines.
How we improved our rank
So, how did we actually improve our ranking?
Traditional SEO tactics actually worked quite well. After determining which prompts we wanted to be included in, we increased keyword implementation across owned, earned, and social media for AI and Venture Capital-focused PR and communications firms by writing a series of blogs relevant to our audience.
We also incorporated a few newer tactics that were more relevant to GEO, including increasing press release distribution and drafting content in FAQ format. Press releases are known to train the models on your messaging repeatedly across multiple sources, so we uploaded a release to the wire about our AI client wins in Q1. The same release was covered in earned media and promoted on social, disseminating that same message even further. For FAQs, the structure of question-and-answer is perfect for training LLMs, knowing this is the format for how AI processes user intent.
As a result, we saw Inkhouse not only maintain its position in certain prompts, but also appear in prompts it wasn’t included in before.
We truly believe this success wasn’t because we were fixated on GEO but because brands that win create an audience-centric strategy across earned, owned, shared and paid distribution channels. Does anyone really know for certain how to influence AI visibility six months from now? Probably not. What’s most important to us is that we continue to tell a compelling story that adds value and keeps our audience genuinely engaged.
